[LLM] Support woq model save and load (#1211)

examples/huggingface/pytorch/text-generation/quantization/run_generation.py

@@ -163,7 +163,7 @@
 # ============AutoModel parameters==============
 parser.add_argument("--load_in_4bit", type=bool, default=False)
 parser.add_argument("--load_in_8bit", type=bool, default=False)
-parser.add_argument("--_commit_hash", default="main", type=str)
+parser.add_argument("--_commit_hash", default=None, type=str)
 parser.add_argument("--trust_remote_code", type=bool, default=False)
 parser.add_argument("--use_llm_runtime", action="store_true")
 # =======================================
@@ -335,22 +335,20 @@
     else:
         user_model = AutoModelForCausalLM.from_pretrained(
             args.model,
-            config=config,
             trust_remote_code=args.trust_remote_code,
             _commit_hash=args._commit_hash,
-            use_llm_runtime=args.use_llm_runtime,
         )
 
 # save model
 if args.output_dir:
+    tokenizer.save_pretrained(args.output_dir)
     if args.sq:
         config.save_pretrained(args.output_dir)
         user_model.save(args.output_dir)
-    elif args.mixed_precision:
-        user_model.config.save_pretrained(args.output_dir)
-        torch.save(
-            user_model.state_dict(), os.path.join(args.output_dir, "pytorch_model.bin")
-        )
+    elif args.mixed_precision or args.woq:
+        user_model.save_pretrained(args.output_dir)
+
+
 
 # int8 model loading
 if args.int8 or args.int8_bf16_mixed:

intel_extension_for_transformers/llm/quantization/nn/modules.py

@@ -115,12 +115,6 @@ def forward(self, x: torch.Tensor):
         # weights are cast automatically as Int8Params, but the bias has to be cast manually
         if self.bias is not None and self.bias.dtype != x.dtype:
             self.bias.data = self.bias.data.to(x.dtype)
-
-        if getattr(self.weight, "quant_state", None) is None:
-            print(
-                "FP4 quantization state not initialized. Please call .quantize_weights()."
-            )
-
         shape = list(x.size())
         m = reduce(mul, shape[0:-1])
         out = torch.zeros(m, self.out_features, dtype=x.dtype)
@@ -143,7 +137,8 @@ def forward(self, x: torch.Tensor):
         return out
 
     def set_weights_bias(self, weight_data, bias=None):
-        shape = weight_data.shape
+        if weight_data.is_meta:
+            weight_data = torch.ones(weight_data.shape, dtype=torch.float)
         weight = torch.ops.bestlaop.woq_quantize(
             weight_data,
             True,
@@ -153,7 +148,6 @@ def set_weights_bias(self, weight_data, bias=None):
             self.scale_dtype if self.scale_dtype is not None else "fp32",
             False,
         )
-        weight.resize_(shape)
         self.weight = ParamsQBits(
             data=weight,
             requires_grad=False,

intel_extension_for_transformers/llm/quantization/utils.py

@@ -198,6 +198,11 @@ def _replace_linear(
                                 module.weight.data,
                                 None if module.bias is None else module.bias.data,
                             )
+                    else:
+                        model._modules[name].set_weights_bias(
+                            module.weight.data,
+                            None if module.bias is None else module.bias.data,
+                        )
                 else:
                     if not hasattr(module, "qweight"):
                         n_pack = 8 // DTYPE_BITS_MAPPING[quantization_config.weight_dtype]

intel_extension_for_transformers/transformers/utils/config.py

@@ -319,6 +319,8 @@ def to_json_file(self, json_file_path: Union[str, os.PathLike], use_diff: bool =
             json_file_path (`str` or `os.PathLike`):
                 Path to the JSON file in which this configuration instance's parameters will be saved.
         """
+        # set tokenizer to None due to it doesn't support write to json
+        self.tokenizer = None
         with open(json_file_path, "w", encoding="utf-8") as writer:
             writer.write(self.to_json_string(use_diff=use_diff))